OpenAI appears to be working on a fix for one of the biggest headaches in desktop AI remote control. According to TestingCatalog, the company is developing a new capability for Codex that would let the AI keep working in the background even when a Mac is locked or asleep.
Until now, all AI screen-control features have had a major flaw: the computer has to be unlocked and awake for the AI to see the screen and simulate mouse clicks and keystrokes. The leaked information suggests OpenAI wants to close that gap. If it works, a user could remotely trigger Codex to run tests or look up data from their phone while they're away from home — without needing to walk back to the machine to physically unlock it. OpenAI is also testing cross-device connectivity, which would let a primary device directly control a Mac Mini running Codex.
Right now, both the existing Codex and rival Claude Code hit the same wall when the system locks. But having an AI try to stay active by bypassing the lock screen password is a direct challenge to macOS's default security model. This kind of aggressive low-level boundary crossing could easily invite Apple's scrutiny — and potentially, its intervention.